Text categorization-assignment of natural language texts to one or more predefined categories based on their content-is an important component in many information organization and management tasks.Different automatic learning algorithms for text categori-zation have different classification accuracy.Very accurate text classifiers can be learned automatically from training examples.

Inductive learning algorithms and representations for text categorization

Within the framework of statistical learning theory we analyze in detail the so-called elastic-net regularization scheme proposed by Zou and Hastie [H. Zou, T. Hastie, Regularization and variable selection via the elastic net, J. R. Stat. Soc. Ser. B, 67(2) (2005) 301-320] for the selection of groups of correlated variables. To investigate the statistical properties of this scheme and in particular its consistency properties, we set up a suitable mathematical framework. Our setting is random-design regression where we allow the response variable to be vector-valued and we consider prediction functions which are linear combinations of elements (features) in an infinite-dimensional dictionary. Under the assumption that the regression function admits a sparse representation on the dictionary, we prove that there exists a particular ''elastic-net representation'' of the regression function such that, if the number of data increases, the elastic-net estimator is consistent not only for prediction but also for variable/feature selection. Our results include finite-sample bounds and an adaptive scheme to select the regularization parameter. Moreover, using convex analysis tools, we derive an iterative thresholding algorithm for computing the elastic-net solution which is different from the optimization procedure originally proposed in the above-cited work.

Elastic-net regularization in learning theory

Disassembly sequencing is important for remanufacturing and recycling used or discarded products. AND / OR graphs (AOGs) have been applied to describe practical disassembly problems by using “ AND ” and “ OR ” nodes. An AOG-based disassembly sequence planning problem is an NP-hard combinatorial optimization problem. Heuristic evolution methods can be adopted to handle it. While precedence and “ AND ” relationship issues can be addressed, OR (exclusive OR ) relations are not well addressed by the existing heuristic methods. Thus, an ineffective result may be obtained in practice. A conflict matrix is introduced to cope with the exclusive OR relation in an AOG graph. By using it together with precedence and succession matrices in the existing work, this work proposes an effective triple-phase adjustment method to produce feasible disassembly sequences based on an AOG graph. Energy consumption is adopted to evaluate the disassembly efficiency. Its use with the traditional economical criterion leads to a novel dual-objective optimization model such that disassembly profit is maximized and disassembly energy consumption is minimized. An improved artificial bee colony algorithm is developed to effectively generate a set of Pareto solutions for this dual-objective disassembly optimization problem. This methodology is employed to practical disassembly processes of two products to verify its feasibility and effectiveness. The results show that it is capable of rapidly generating satisfactory Pareto results and outperforms a well-known genetic algorithm.

Modeling and Planning for Dual-Objective Selective Disassembly Using and / or Graph and Discrete Artificial Bee Colony

We propose an embedded/integrated feature selection method based on neural networks with Group Lasso penalty. Group Lasso regularization is considered to produce sparsity on the inputs to the network, i.e., for selection of useful features. Lasso based feature selection using a multi-layer perceptron usually requires an additional set of weights, while our Group Lasso formulation does not require that. However, Group Lasso penalty is non-differentiable at the origin. This may lead to oscillations in numerical simulations and make it difficult to analyze theoretically. To address this issue, four smoothing Group Lasso penalties are introduced. A rigorous proof for the convergence of the proposed algorithm is presented under suitable assumptions. To verify the effectiveness, a three-step algorithmic architecture is adopted in implementation. Experimental results on several datasets validate the theoretical results and demonstrate the competitive performance of the proposed method.

Feature Selection for Neural Networks Using Group Lasso Regularization

A Bolasso based consistent feature selection enabled random forest classification algorithm: An application to credit risk assessment

High-order covariate interacted Lasso for feature selection

This paper presents an education framework to effectively develop crucial software engineering skills in students of software Engineering major at National Exemplary Software School (NESS), Xiamen University. The goal is to describe a systematic approach towards integrating project based learning in software engineering major, both inside and outside the classroom. An essential part of Software Engineering Education is practical training in principles, methods and procedures under conditions similar to developing real software products. This paper describes the different conventional and traditional approaches at length for Software Engineering Education and proposes integrated project based learning approach is more effective and interesting for teaching and learning SE as compared to the lecture based approach.

Integrated project based learning in software engineering education

As an important part of traditional medicine, TCM (Traditional Chinese Medicine) has unique and distinct clinical effects in the aspect of disease diagnosis and treatment. Thousands of years of TCM treatment has accumulated abundant clinical data and medical literatures, including valued TCM theories and clinical practice rules. Researchers have conducted various methods such as clustering analysis, association rules and regression analysis to study TCM theory. However, none of them could reflect well the semantic complexity of TCM and systemic characteristics of TCM treatment. This paper conducted a research on the inherent rules of TCM clinic records with a topic model. On the basis of LDA model, weighted mechanism was adopted for each feature word to improve the distinguishing ability and interpretability between the topics. Meanwhile, the modeled topic is taken as the feature for the classification of SVM (Support Vector Machine) to improve the classification accuracy. The topic number of LDA topic model is confirmed by the KL distance and similarity between the topics. After analyzing the relationship between topic model and TCM differentiation and treatment, MULTI-RELATIONSHIP Topics LDA MODEL was proposed on the basis of LDA model and Author-topic model to automatically extract the topic structures between the four parties and explore the relationship of the multiple parties with clinical significance. In the meantime, relevancy between the parties and the feature word weighted mechanism are used to improve the MULTI-RELATIONSHIP Topics LDA MODEL and the classification accuracy of the topics. The experiments showed that analysis of clinical data with topic model can extract TCM treatment rules and provide a novel theoretical method for TCM clinical research.

TCM clinic records data mining approaches based on weighted-LDA and multi-relationship LDA model

In this paper, we investigate the problem of optimizing complex multivariate performance measures to learn classifiers for pattern classification problems. For the first time, the multi-kernel learning is considered to construct a classifier to optimize a given nonlinear and non-smooth multivariate classifier performance measure. We estimate and optimize the upper bound of the given multivariate performance measure, instead of optimizing it directly. Moreover, to solve the problem of kernel function selection and kernel parameter tuning,
 we proposed to construct an optimal kernel by weighted linear combination of some candidate kernels. The learning of the classifier parameter and the kernel weight are unified in a single objective function considering minimizing the upper bound of the given multivariate performance measure. The objective function is optimized with regard to classifier parameter and kernel weight alternately in an iterative algorithm. The developed algorithm is evaluated on two different pattern classification methods with regard to various multivariate performance measure optimization problems. The experiment results show the proposed algorithm outperforms the competing methods.

Multi-kernel learning for multivariate performance measures optimization

The human eye's state of motion and content of interest can express people's cognitive status and emotional status based on their situation. When observing the surrounding things, the human eyes make different eye movements according to the observed objects which reflects human's attention and interest. In this paper, we capture and analyze patterns of human eye-gaze behavior and head motion and classify them into different categories. Besides, we compute and train the eye-object movement attention model and eye-object feature preference model based on different peoples' eye-gaze behaviors by using machine learning algorithms. These models are used to predict humans' object of interest and the interaction intention according to people's real-time situation. Furthermore, the eye-gaze behavior and head motion patterns can be used as a modality of non-verbal information in the computing of human emotional states based on the PAD affective computing model. Our methodology analyzes human emotion and cognition status from the aspect of eye-gaze behavior and head motion, understands the cognitive information that human eyes can express, and effectively improves the efficiency of human-computer interaction in different circumstances.

Jianbing Xiahou

Papers

High-order covariate interacted Lasso for feature selection

Integrated project based learning in software engineering education

TCM clinic records data mining approaches based on weighted-LDA and multi-relationship LDA model

Multi-kernel learning for multivariate performance measures optimization

Real-Time Eye-Gaze Based Interaction for Human Intention Prediction and Emotion Analysis